These guys have no idea what they are talking about.

[Cyber-Mav]

Are you saying alderlake is optimised for benchmarks and not real world application?

 

[Bencher]

Fair enough.



There is more to assessing a core's multithread throughput than just cinebench.

On the criteria of MT performance, they are clearly not equivalent to Zen 3, as is evidenced by all the multithreaded tests in that same TPU review, including several other rendering benchmarks, not to mention other highly parallel workloads also shown.
You are very focused on a benchmark where the e-cores are particularly good at maximising instruction throughput for that workload to prove a conclusive point, but which is not indicative of real world multi thread performance or efficiency.

Speaking of cinebench, what would be far more interesting than a loaded comparison using a preferred favourable benchmark, is why exactly does cinebench favour the e-core.

Well sure, the ecores work better in cache agnostic avx workloads I think, but you are missing the point. The point is they are coprocessors used in mt workloads, it doesnt matter if they are not that fast in whichever other application (wprime for example) cause you already have the Pcores for those cases

 

[Cyber-Mav]

And raptorlake is gonna double up on e cores?
Will they be still gracemont or upgraded?

 

[alec]

And raptorlake is gonna double up on e cores?
Will they be still gracemont or upgraded?

current rumors say

Little cores are Gracemont, same as Alder Lake, just with twice the cache

Doubling up to 16 small cores only for top 13900 chips, the rest get up to 8

 

[Cyber-Mav]

Any update on the rumours of the e core cpu only that has 96 cores?

 

[keyser van someone]

Any update on the rumours of the e core cpu only that has 96 cores?

I'm so down with the kids.